Ankylosing spondylitis and blood tests

What the research found.

• One Lebanese study tested 141 people with axial spondyloarthritis and 106 blood donors. HLA-B27 was present in 41.1 percent of patients and in 3.8 percent of donors. As a test it caught 41.1 percent of cases and cleared 96.2 percent of donors, with a positive likelihood ratio of 10.9 and a negative one of 1.63.

  Ziade and colleagues, International Journal of Rheumatic Diseases, 2019

• One study compared the available tests head to head against each other. A polygenic risk score reached an area under the curve of 0.924 in people of European descent, while HLA-B27 alone reached 0.869, a sacroiliac MRI 0.885, and C-reactive protein 0.700. A score built in people of East Asian descent reached 0.948.

  Li and colleagues, Annals of the Rheumatic Diseases, 2021

• One study across 24 countries typed 2269 people with axial spondyloarthritis, of whom 1753 were positive and 516 negative. Positive patients were diagnosed younger, at a mean of 31.6 years against 37.7 years. They were more often men, at 72.1 percent against 54.3 percent.

  Chaudhary and colleagues, RMD Open, 2023

• One study scanned 102 people with sudden inflammation at the front of the eye, alongside 39 healthy people for comparison. Bone marrow edema is fluid visible on MRI inside the bone itself, and it was found in the sacroiliac joints of 51 percent of the patients. It was also found in 28 percent of the healthy group, and that is the single figure on this page most worth reading twice over.

  Bubova and colleagues, Diagnostics, 2022

• One analysis of 925 people with ankylosing spondylitis compared three ethnic groups against each other. The median sedimentation rate was 27.0 in Black patients, 10.0 in White patients, and 17.0 in Latino patients, at P under 0.0001. Disease activity and C-reactive protein differed in the identical direction.

  Jamalyaria and colleagues, Clinical Rheumatology, 2017

What the number means.

No blood test makes this diagnosis

People come to this subject expecting one test to settle it, and there isn't one. What there is instead is a set of results, each of which moves the odds by an amount somebody has measured. Knowing how far each result moves them is most of what this page is for.

HLA-B27 is the test everybody has heard of, and it's a gene marker found by an ordinary blood test. It's strongly associated with this disease, and plenty of well people have it too. Those two facts have to be held together at once, and that is what makes this test hard to interpret.

So the useful question isn't whether the test comes back positive. It's how far a positive result moves things, how far a negative one does, and what else is being weighed alongside them both. Those three questions are what the rest of this page answers.

What HLA-B27 does as a test

One Lebanese study answered that question properly, by testing 141 people with axial spondyloarthritis and 106 blood donors drawn from that population. It then reported the full set of test performance figures, which most papers don't. HLA-B27 was present in 41.1 percent of patients and in 3.8 percent of donors.

So as a test it caught 41.1 percent of cases and correctly cleared 96.2 percent of people without the disease. Two more figures follow directly from those two. The positive likelihood ratio was 10.9, which means a positive result moves the odds a long way, and the negative likelihood ratio was 1.63, which barely moves them at all.

Read those together and the picture is a clear one. This test is strong when it's positive and weak when it's negative, which is a common thing in medicine. It's also the opposite of what most people assume about a genetic test.

Why a negative result means different things in different places

How often the gene turns up varies enormously by ancestry, and that changes what a negative result is worth. It changes it by a very large amount. The studies below are the reason this page keeps naming where each one was done.

One Moroccan study covered 136 people with this family of diseases, and only 7 of them had the gene at all. In that population the test would miss almost everybody who had the disease. One Kurdish study found the gene in 65.9 percent of 41 patients and in 3.8 percent of 209 healthy donors.

One Singaporean study found it in 82 percent of those with established disease and 68 percent of those with the earlier form. So a negative result holds very different weight in Rabat, Beirut, Singapore, and Erbil. A team that knows this reads your test differently, which makes it a fair thing to ask about.

What an area under the curve means

One study ranked four available tests against each other, asking how well each of them separated people who had ankylosing spondylitis from people who didn't, using a scale called the area under the curve. A perfect test scores 1.0 on that scale, and a test that's no better than tossing a coin scores 0.5. So a score near 1.0 is a strong discriminator and a score near 0.5 is a weak one, which is the whole of what the four numbers below say.

A polygenic risk score, meaning many gene variants scored together rather than one gene read on its own, did best of the four: 0.924 in people of European descent and 0.948 in people of East Asian descent. It's a research tool rather than something your team can order today, so it's a development to watch rather than a question to raise at your next appointment. The other three tests followed along behind it in a fairly tight group, with a sacroiliac MRI at 0.885, HLA-B27 alone at 0.869, and C-reactive protein at 0.700.

C-reactive protein sits a good deal closer to the bottom of that range than most people assume it does. That's worth knowing when somebody tells you your inflammation marker came back normal, because normal there means less than it sounds. One more figure from that study is worth taking away with you.

It assumed a prior probability of 10 percent, which is roughly the situation in chronic back pain under the age of 45. On that footing the polygenic score's best positive predictive value was 78.2 percent. So the best available score, used in the setting it was built for, still leaves about one in five wrong.

No test anywhere on this page does better than that figure does. It's worth holding onto while you read the rest of the numbers here, because they're all weaker than that one. A test that's right four times out of five is still wrong the fifth time.

What C-reactive protein is still good for

That doesn't mean the test is useless to anybody. It means the test is poor at telling who has this particular disease and who doesn't. Telling who has a disease isn't the only job a blood test is asked to do.

One study followed 819 people with ankylosing spondylitis, looking for a complication at the top of the neck. That complication was found in 14.1 percent of them. A raised C-reactive protein was one of three things independently associated with developing it, at an odds ratio of 2.19 from 1.36 to 3.53.

Peripheral arthritis came in at 2.05, on an interval from 1.36 to 3.07. One larger study of 1075 patients found that direction repeated in its own data. Those with peripheral arthritis on examination had higher sedimentation rates and higher C-reactive protein, both of them at P under 0.001.

So a raised result is worth following up on. It just isn't the thing that tells anybody whether you have this disease in the first place. Those are two different jobs, and a test can be good at one of them and poor at the other.

The scan, and the healthy people in it

One study scanned 102 people with sudden inflammation at the front of the eye. Bone marrow edema, which is fluid visible on MRI inside the bone rather than in the joint space around it, was found in the sacroiliac joints of 52 of them, which comes to 51 percent. Edema highly suggestive of axial spondyloarthritis was found in 33 of them, which is 32 percent, and that is the finding the page rests on.

The comparison group is the part of that study worth reading twice over. It scanned 39 healthy people, and bone marrow edema turned up in 11 of them, which is 28 percent, though not one of them had the highly suggestive kind. Chronic back pain was equally common in both of the groups.

That's another reason the symptom on its own sorts you into nothing. Forty of those 102 patients met the formal classification criteria for axial spondyloarthritis. So a scan finding is information rather than an answer, and it needs the rest of the picture sitting beside it.

More than a quarter of well people had something on that scan. Reading it without the rest of the picture would have labelled them with something they didn't have. That's the argument against a scan resting on its own, and it rests on a comparison group of 39.

What your HLA-B27 status changes about the picture

One study across 24 countries typed 2269 people with axial spondyloarthritis, of whom 1753 were positive and 516 were negative. It then compared the two groups against each other. Positive patients were diagnosed younger, at a mean of 31.6 years against 37.7 years.

They were more often men, at 72.1 percent against 54.3 percent, and a positive family history was commoner too, at 29.8 percent against 15.3 percent. Negative patients had more going on outside the spine. Peripheral arthritis was commoner in them, at 47.5 percent against 42.1 percent, and so was psoriasis, at 19.4 percent against 10.2 percent.

Read all of that as a description, not as a ranking, because neither group is the severe one. They're two somewhat different illnesses wearing one name between them. Knowing which of them you have helps your team know what to watch for.

Numbers are read against people

One analysis compared 925 people with ankylosing spondylitis across three ethnic groups, and it found differences in almost everything it measured. The blood tests were only part of that. The median sedimentation rate was 27.0 in Black patients, 10.0 in White patients, and 17.0 in Latino patients.

That difference came in at P under 0.0001, and C-reactive protein differed in the identical direction. So did the rest of what the study measured. Median functional impairment was 62.5 against 27.8 and 38.1, median disease activity was 5.9 against 3.5 and 4.5, and baseline X-ray scores were higher too.

This page won't offer an explanation for that, because the study doesn't give one and every available explanation is disputed. What it does mean is practical enough. Your result is read against you and your own history rather than against an average, and a team that knows this reads it better.

One thing worth mentioning quickly

Most of this page is about not over-reading a number, and this part is the opposite of that. It's the one place on the page where acting fast counts for something. One study covered 301 people with this family of diseases.

Of them, 82, or 27.2 percent, had at least one episode of inflammation inside the eye. Prevalence was 11.5 percent at diagnosis and rose to 39.3 percent twenty years later. HLA-B27 positivity was independently associated with it, at a hazard ratio of 4.5, on an interval ranging from 1.3 to 15.2.

That interval is wide enough that the direction is the part to read rather than the size. Sudden eye pain, redness, or blurred vision needs seeing the day it happens. So does new trouble with light, and that's the one thing on this page where waiting costs something that can't be recovered.

Common misconceptions.

Cautions specific to this condition.

• Don't take a negative HLA-B27 as an all-clear. It caught 41.1 percent of cases in one Lebanese study. Only 7 of 136 patients had it in one Moroccan study.
• Don't take a positive HLA-B27 as a diagnosis either. It was present in 3.8 percent of healthy blood donors in that Lebanese study.
• Ask what your C-reactive protein is being used for. It separated patients from controls at 0.700, where 0.5 is a coin toss and 1.0 is perfect.
• Tell your team about eye symptoms. HLA-B27 positivity predicted uveitis at a hazard ratio of 4.5, from 1.3 to 15.2, in one study of 301 people.
• A raised C-reactive protein is worth following up. It was associated with a neck complication at an odds ratio of 2.19, from 1.36 to 3.53, in one study of 819 people.
• This page gives no reference range and no target for any test. Ranges differ by laboratory, and what your result means is your team's call.

Questions patients ask.